In November 2022, a developer named Guillaume Luccisano sat down and built a customer support bot "for fun." He dropped it onto the Gorgias App Store to see what would happen. What happened was an inbox flood of demo requests. He had built something merchants actually wanted. Days later, OpenAI released ChatGPT to the public, and the world caught up to what Guillaume had already seen.
That prototype became Yuma AI. Officially launched in December 2022, accepted into Y Combinator's Winter 2023 batch, and now the support automation infrastructure behind more than 100 ecommerce brands - including Glossier, EvryJewels, and Clove. The company's thesis is simple enough to fit on a napkin: most customer support tickets are repetitive, and repetitive tasks should not require humans.
The Man Who's Done This Before (Twice)
Guillaume Luccisano is not a first-time founder nervously navigating Series A anxiety. He is a three-time Y Combinator founder, which places him in a category so small you could fit it in a support queue. His first company, Socialcam, graduated YC W12 and was acquired by Autodesk for $60 million. His second, Triplebyte - a developer hiring platform - came out of YC S15, raised $50 million, and was later acquired by Karat. His third is Yuma, and based on the investor roster who followed him in, the market has learned to pay attention when he moves.
"Help merchants optimize their support operations through advanced automation while maintaining a disciplined approach to growth."- Guillaume Luccisano, Founder & CEO, Yuma AI
There is a pattern here worth noting. Socialcam was early at Twitch. Triplebyte was early on the "screen the developer before the interview" idea. Yuma was early on AI support agents - built before ChatGPT, pitched before the world was ready for the vocabulary. The timing was not luck; it was the habit of someone who watches platforms while everyone else watches products.
What Yuma Actually Does
The product does not replace a helpdesk. It installs inside the helpdesks merchants already use - Gorgias, Zendesk, Kustomer, Salesforce Service Cloud - and intercepts tickets before a human touches them. The AI reads the ticket, pulls relevant data from Shopify or BigCommerce, and resolves it: issues the refund, initiates the return, answers the WISMO ("where is my order?") query, modifies the subscription. If it cannot, it escalates with full context.
Zero engineering is required to deploy. That is not a marketing footnote - it is a product philosophy decision. Ecommerce CX teams are lean, non-technical, and allergic to six-week integration projects. Yuma deploys in days, starts on 5-10% of tickets, and scales. Merchants see automation rates between 60% and 93% of all conversations. At that level, a 10-person support team becomes a 2-person support team with the same output and faster response times.
How the Safety Guard Works
Every AI response passes through 15-20 multilayer checks before delivery. Refund caps and discount ceilings are hard-coded - the AI cannot exceed merchant-defined limits regardless of how politely a customer asks. 180+ languages supported, including RTL scripts and regional dialects. SOC 2 Type II certified as of Q1 2026.
The Product Suite
Yuma has grown from a single support tool into a four-channel commerce AI platform. Support was the entry point, but the company's ambition is the full customer interaction layer.
Support AI
Autonomous ticket resolution inside existing helpdesks. Handles returns, refunds, WISMO, subscriptions, and order changes end-to-end.
Sales AI
On-page assistant for Shopify product pages. Real-time pre-purchase Q&A - reported 6% conversion lift, 18% higher revenue per visitor.
Social AI
Automates Facebook, Instagram DMs, TikTok Shop, Twitter/X, and review platforms (Trustpilot, Yotpo, Judge.me). Launched January 2026.
Chat AI
AI-native live chat widget. Handles product, policy, and order questions; escalates to helpdesk with full context when needed.
Ask Yuma
Merchants upload SOPs; the AI reads them, asks clarifying questions, and generates deployable automation flows. Launched April 2026 with Claude MCP integration.
Flows + Deep Search
Visual workflow builder for structured automations, and natural language querying across all ticket history. Both launched November 2025.
The Numbers Don't Need a Spin Cycle
Customer results are specific enough to be credible and varied enough to be real. EvryJewels handles 150,000 monthly tickets; Yuma automates 89% of them, dropping cost per ticket from $5.50 to $2.00. Glossier, a brand that built its entire identity on customer intimacy, hits 91% accuracy on complex tickets and 87% faster first-response time. Clove went from a one-day first-response time to three minutes. Tediber, a French mattress brand, cut response time from 72 hours to under one hour.
| Customer | Automation | Key Result |
|---|---|---|
| EvryJewels | 89% | $5.50 → $2 cost/ticket, 150K monthly tickets |
| Glossier | 91% | 87% faster first-response time |
| Clove | 70% | FRT: 1 day → 3 minutes, 3x ROI |
| CABAIA (France) | 74% | 74% cost reduction (€3.75 → €1 per ticket) |
| Tediber (France) | 64% | Response time: 72 hours → under 1 hour |
Who Backed Them
The $5 million seed round closed in October 2024, led by Google's Gradient Ventures. Altman Capital Fund, Karman Ventures, and AI Grant joined alongside a roster of individual investors that reads like a who's-who of founder-turned-angels: Kyle Vogt (Cruise), Emmett Shear (Twitch), Nicolas Dessaigne (Algolia), Garry Tan (Y Combinator), Solomon Hykes (Docker), and Justin Kan (Twitch). When serial founders invest in a serial founder, they are not betting on the idea - they are betting on the pattern.
The most unusual investor is Motier Ventures, the venture arm of Galeries Lafayette - the Parisian department store founded in 1893. A 130-year-old French luxury retailer backing a 2-year-old Boston AI startup is either a sign of strategic foresight or proof that everyone in retail now lies awake thinking about their support queue costs.
The Business Model That Changed Everything
In September 2024, Yuma did something that most SaaS companies only discuss in board decks: they switched to outcome-based pricing. Merchants pay per resolved ticket, not per seat or per month. If Yuma does not solve the problem, Yuma does not get paid. The move is either bold confidence in the product's accuracy or the clearest possible signal that they already know the automation rate is high enough to make it work. Given the customer data, it appears to be both.
Pricing Philosophy
Outcome-based model launched September 2024. Merchants pay approximately $0.65-0.70 per resolved ticket. 30-day free trial on live helpdesk data. No engineering required. Dedicated account manager and private Slack channel included at every tier. If Yuma doesn't resolve the ticket, you don't pay for it.
The Road Forward
The product roadmap for 2025 and 2026 reveals where Guillaume sees the ceiling - or rather, that he does not see one. The company launched Flows (visual workflow automation builder) and Deep Search (natural language querying across ticket history) in November 2025. Social AI - covering Facebook, Instagram, TikTok Shop, and Twitter/X - arrived in January 2026. Ask Yuma 2.0, a conversational interface that turns merchant SOPs into deployable automations, launched in April 2026 with a Claude MCP integration.
The pattern is horizontal expansion: from email tickets, to chat, to social, to the full commerce conversation surface. Each new channel is another revenue stream, another automation rate, and another argument that Yuma is building infrastructure rather than a point solution.